ARTICLE IN PRESS

European Economic Review 52 (2008) 1–27

www.elsevier.com/locate/eer

Terrorism and the world economy

Alberto Abadiea, Javier Gardeazabalb,
a
John F. Kennedy School of Government, Harvard University and NBER, 79 John F. Kennedy Street,
Cambridge, MA 02138, USA
b
Dpto. Fundamentos del Análisis Económico II, University of the Basque Country, Avda. Lehendakari Aguirre 83,
48015 Bilbao, Spain
Received 27 March 2007; accepted 20 August 2007
Available online 9 October 2007

Abstract

It has been argued that terrorism should not have a large effect on economic activity, because
terrorist attacks destroy only a small fraction of the stock of capital of a country (see, e.g., Becker,
G., Murphy, K., 2001. Prosperity will rise out of the ashes. Wall Street Journal October 29, 2001). In
contrast, empirical estimates of the consequences of terrorism typically suggest large effects on
economic outcomes (see, e.g., Abadie, A., Gardeazabal, J., 2003. The economic cost of conflict:
A case study of the Basque country. American Economic Review 93, 113–132). The main theme of
this article is that mobility of productive capital in an open economy may account for much of the
difference between the direct and the equilibrium impact of terrorism. We use a simple economic
model to show that terrorism may have a large impact on the allocation of productive capital across
countries, even if it represents a small fraction of the overall economic risk. The model emphasizes
that, in addition to increasing uncertainty, terrorism reduces the expected return to investment. As a
result, changes in the intensity of terrorism may cause large movements of capital across countries if
the world economy is sufficiently open, so international investors are able to diversify other types of
country risks. Using a unique data set on terrorism and other country risks, we find that, in
accordance with the predictions of the model, higher levels of terrorist risks are associated with lower
levels of net foreign direct investment positions, even after controlling for other types of country
risks. On average, a standard deviation increase in the terrorist risk is associated with a fall in the net
foreign direct investment position of about 5% of GDP. The magnitude of the estimated effect is

Corresponding author. Tel.: +34 946013780.

E-mail addresses: alberto_abadie@harvard.edu (A. Abadie), javier.gardeazabal@ehu.es (J. Gardeazabal).

0014-2921/$ - see front matter r 2007 Elsevier B.V. All rights reserved.
doi:10.1016/j.euroecorev.2007.08.005
 ARTICLE IN PRESS
2 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

large, which suggests that the ‘‘open-economy channel’’ impact of terrorism may be substantial.
r 2007 Elsevier B.V. All rights reserved.

JEL classification: F21; F41; H56

Keywords: Foreign direct investment; Terrorism; Capital movements; Portfolio allocation

1. Introduction

This paper analyzes the effects of terrorism in an integrated world economy. From an
economic standpoint, terrorism has been described to have four main effects (see, e.g., US
Congress, Joint Economic Committee, 2002). First, the capital stock (human and physical)
of a country is reduced as a result of terrorist attacks. Second, the terrorist threat induces
higher levels of uncertainty. Third, terrorism promotes increases in counter-terrorism
expenditures, drawing resources from productive sectors for use in security. Fourth,
terrorism is known to affect negatively specific industries such as tourism.1 However, this
classification does not include the potential effects of increased terrorist threats in an open
economy. In this article, we use a stylized macroeconomic model of the world economy
and international data on terrorism and the stock of foreign direct investment (FDI) assets
and liabilities to study the economic effects of terrorism in an integrated world economy.
The motivation to study the impact of terrorism in an open world economy is the
following. It has been documented that the direct impact of terrorist attacks on productive
capital is relatively modest. This seems to be true even for events of catastrophic terrorism.
For example, Becker and Murphy (2001) estimated that the September 11th terrorist
attacks resulted in a loss of 0.06% of the total productive assets of the US economy. In
consequence, after taking into account the four channels mentioned in the previous
paragraph, some authors have argued that terrorism is unlikely to exert a significant
influence on economic activity in the long run. The calculations in Becker and Murphy
(2001) bound the long-run effect of the September 11th attacks to 0.3% of GDP (see also
IMF, 2001a and OECD, 2001).2
In contrast, reduced-form estimates of the economic effects of terrorism typically
suggest much larger effects, at least in those areas where the risk of terrorism is particularly
severe or sustained. For example, in our previous study of the impact of terrorism in the
Basque Country, we find a 10% drop in per capita GDP which emerges during a period of
two decades and that is attributable to the terrorist conflict (Abadie and Gardeazabal,
2003). Chen and Siems (2004), Enders and Sandler (1996), and Pshisva and Suarez (2006),
among others, similarly find large effects of terrorism on economic variables.3 However, as
noted by Becker and Rubinstein (2004), the question of why terrorism may have a large
effect on the economy, even if it represents a small fraction of the total economic risk, has
attracted much less attention in the academic literature.

1
See Enders et al. (1992) on the effect of terrorism on tourism.
2
In a more recent paper, Becker and Rubinstein (2004) have argued that terrorism risk may have a large
economic impact if the fear of terrorism affects individual utility in each state of nature.
3
In related research, Frey et al. (2004) study the effect of terrorism on life satisfaction. Frey et al. (2007) surveys
the existing research on the economic impact of terrorism.
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 3

The main theme of this paper is that mobility of productive factors in an open economy
may account for much of the difference between the direct effect and the equilibrium effect
of terrorism on the economy. If terrorism is a local phenomenon, capital will tend to flow
to destinations without a terrorist threat, reducing net foreign investment in the economies
affected by terrorism. Even if terrorism is a global threat, international investment will
respond to differences in the expected intensity of terrorism across countries. In fact,
because the optimal allocation of capital across countries depends not only on the level of
terrorism but also on other country factors that affect the distributions of the returns to
capital, variations in the overall level of terrorism in the world may induce a re-allocation
of capital across countries even if the relative intensity of terrorist risk across countries
remains unchanged.
The amounts of FDI in the US before and after the September 11th attacks provide
some suggestive evidence of the open-economy channel of terrorism. In the year 2000, the
year before the terrorist attacks, FDI inflows represented about 15.8% of the Gross Fixed
Capital Formation in the US. This figure decreased to only 1.5% in 2003, two years after
the attacks. Conversely, FDI outflows from the US increased from about 7.2% of the
Gross Fixed Capital Formation for the US in 2000 to 7.5% in 2003 (see UNCTAD, 2004).
Of course, not all this variation in FDI can be attributed to the effect of the September
11th attacks. As of September 2001 FDI inflows had fallen from its 2000 peak not only in
the US but also in other developed economies (see UNCTAD, 2002). These figures,
however, motivate the question of to which extent an increase in the perceived level of
terrorism was responsible for the drop in FDI in the US that followed the events of
September 11th.
Surveys of international corporate investors provide direct evidence of the importance of
terrorism on foreign investment. Corporate investors rate terrorism as one of the most
important factors influencing their FDI decisions (see Global Business Policy Council,
2004).
To illustrate the importance of the ‘‘open-economy channel’’ of terrorism we
use a stochastic version of the AK endogenous growth model (see, e.g., Obtsfeld,
1994; Turnovsky, 1997). We extend this model by introducing terrorism as a
stochastic Poisson process, with events that destroy some fraction of the capital stock of
a country.
The model emphasizes that, beyond increasing uncertainty, terrorism reduces the
expected return to investment. As a result, changes in the intensity of terrorism have an
ambiguous effect on the overall investment position of the world (investments over
wealth), but they may cause large movements of capital across countries if the world
economy is sufficiently open, so international investors are diversified against other types
of country risks.
One of the predictions of our model is that, like any other risk, terrorism should affect
the stock of international investment in any particular country. Therefore, it is possible to
obtain empirical evidence on the ‘‘open-economy channel’’ of terrorism by looking at the
relationship between the stock of net foreign investment and terrorism in the cross-section
of countries, as long as we account for other factors that affect international investment
positions, particularly other country risks which may be correlated with terrorism levels.
For this purpose, we use a unique international data set on terrorism risk and other types
of country risks. We find that terrorism has a negative and sizeable impact on foreign
investment positions.
 ARTICLE IN PRESS
4 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

Enders and Sandler (1996) have also studied the effect of terrorism on capital
flows across countries. Using vector autoregression methods, these authors estimate a
negative 13.5% effect of terrorism on FDI for Spain (for the period 1976–1991)
and a negative 11.9% effect for Greece (for the period 1975–1991). Our empirical
results for a cross-section of countries corroborate and provide external validity to
the results of Enders and Sandler. In addition, Enders et al. (2006) find that terrorist
attacks against US interests in OECD countries significantly reduced stocks of US direct
investment. In related research, Blomberg and Mody (2005), report a significant effect of
violence (including terrorism) on the inflow of direct investment in a sample of 43
countries.
The rest of the article is organized as follows. In Section 2, we built a simple model that
illustrates why terrorism may have a large effect on net foreign investment in an open
world economy even if terrorism induces only a small fraction of the total economic risk.
Section 3 describes the data set. Section 4 provides empirical evidence on the effect of
terrorism on FDI. Section 5 concludes.

2. A simple model of catastrophic terrorism

2.1. The model

Consider a two-country economy with terrorism and perfect capital mobility across
countries. We will refer to one of the countries as the ‘‘domestic economy’’ and to the other
as the ‘‘foreign economy’’. The world population consists of a continuum of identical and
infinitely lived agents with mass equal to one, who are equally distributed among the two
countries. At each point in time, t, agents decide how much to consume, CðtÞ, and which
fraction, vðtÞ, of the capital to devote to production in the domestic economy (with a
fraction 1  vðtÞ devoted to production in the foreign economy). If the fraction of capital
devoted to a country changes, this change generates a flow of investment from one country
to the other.
As in Obstfeld (1994) and Turnovsky (1997), we assume that production in the domestic
economy is given by a stochastic AK technology:
dY ðtÞ ¼ avðtÞKðtÞ dt þ sW vðtÞKðtÞ dW ðtÞ,
where dY ðtÞ is output, KðtÞ is the world stock of capital (physical and human), and W ðtÞ is
a Wiener process, whose innovations capture domestic productivity shocks.
The assumptions of constant returns to scale and perfect capital mobility across
countries are not totally innocuous. These assumptions increase the sensitivity of the
allocation of capital across countries to differences in the distributions of the return to
capital between countries. The assumptions of constant returns to scale and perfect capital
mobility across countries are likely to be violated in the short run. However, these
assumptions are consistent with the long-run trends in the allocation of capital across
countries.4
4
McGrattan (1998) and Li (2002) present evidence that long-run trends in investment and growth are consistent
with the predictions of the AK model. More importantly for the purpose of this article, Kraay and Ventura (2000,
2002) show that the observed long-run patterns in the allocation of capital across countries are consistent with
weak diminishing returns to capital.
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 5

Terrorist attacks in the domestic economy are captured in this model as innovations
from a Poisson process, PðtÞ with rate l, which destroy a fraction d of the stock of capital
allocated by every investor to the domestic economy, with 0pdp1. After the direct impact
of terrorism is taken into account, the return to capital in the domestic economy is
governed by a jump-diffusion:
dY ðtÞ  dvðtÞKðtÞ dPðtÞ
dRðtÞ ¼ ¼ a dt þ sW dW ðtÞ  d dPðtÞ.
vðtÞKðtÞ
By the properties of Wiener and Poisson processes, the expectation and variance of the
return to capital in the domestic economy are
E½dRðtÞ ¼ ða  ldÞ dt,
and

varðdRðtÞÞ ¼ ðs2W þ ld2 Þ dt.

Because terrorism is a one-sided risk (that is, because it produces negative shocks only), an
increase in the intensity of domestic terrorism, l, not only increases the variance of the
return to capital but it also reduces its mean. The one-sided risk nature of terrorism is
crucial to derive the results below.
Production and terrorism follow analogous processes in the foreign economy, for which
we will use the notation: a , sW , d , l , dW  ðtÞ, and dP ðtÞ.
Agents derive instantaneous utility from consumption, CðtÞ, through a constant relative
risk aversion utility function: uðcÞ ¼ ðc1g  1Þ=ð1  gÞ, with g40 and c40 (taking the
limiting form uðcÞ ¼ lnðcÞ, for g ¼ 1). The parameter g is the Arrow–Pratt measure of
relative risk aversion, g ¼ c u00 ðcÞ=u0 ðcÞ.
Agents choose CðtÞ and vðtÞ to maximize lifetime discounted utility, subject to the law of
motion for capital. Because all agents have the same preferences and investment
possibilities, regardless of how ownership of productive capital is distributed, the
equilibrium in the world economy is given by the solution of the utility maximization
problem for a representative agent:
Z 1 
CðtÞ1g  1
max E ebt dt
0 1g
s:t: dKðtÞ ¼ ðavðtÞKðtÞ þ a ð1  vðtÞÞKðtÞ  CðtÞÞ dt
þ sW vðtÞKðtÞ dW ðtÞ þ sW ð1  vðtÞÞKðtÞ dW  ðtÞ
 dvðtÞKðtÞ dPðtÞ  d ð1  vðtÞÞKðtÞ dP ðtÞ,
0pCðtÞpKðtÞ; KðtÞX0; Kð0Þ ¼ K 0 ; 0pvðtÞp1. ð1Þ

Appendix A provides a detailed derivation of the solution. The optimal consumption

plan is

1 1
CðtÞ ¼ v þ a ð1  b
b þ ðg  1Þðab vÞÞ  gðg  1Þðs2W bv2 þ s2 vÞ2 Þ
W ð1  b
g 2

vÞ1g  1  l ½ð1  d ð1  b
 l½ð1  db vÞÞ1g  1 KðtÞ, ð2Þ
 ARTICLE IN PRESS
6 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

where bv is the optimal share of world capital invested in the domestic economy, which is
implicitly determined by
ða  a Þ  gðs2W b
v  s2
W ð1  b vÞg þ l d ð1  d ð1  b
vÞÞ  ldð1  db vÞÞg ¼ 0. (3)
Notice that Eq. (3) implies that the optimal share of capital invested in the domestic
economy is constant, for any given values of the parameters of the model.
Let y be the fraction of the world’s productive capital owned by residents of the
domestic economy. Domestic consumption and wealth are equal to yCðtÞ and yKðtÞ,
respectively. Similarly, foreign consumption and wealth are equal to ð1  yÞCðtÞ and
ð1  yÞKðtÞ. In this economy, domestic and foreign residents hold the same portfolio of
assets, a share in the world portfolio. Hence, the distribution of consumption and wealth
among countries depends only on the value of y. However, the distribution of the stock of
capital between countries depends on the intensity of terrorism in both countries and the
other parameters of the model.

2.2. The effects of terrorism

In this economy, terrorism affects capital accumulation through three different

channels. First, terrorist events directly destroy part of the capital stock of a country, d.
As explained above, in practice, the quantitative importance of this effect seems to be
small.
Second, terrorism changes the process that determines the return to capital, affecting the
overall investment position of the individuals in the world economy. However, the
direction of this second effect is theoretically ambiguous. In the absence of a terrorist
attack, every unit of capital is either consumed or saved as productive capital. Let pC ¼
CðtÞ=KðtÞ be the consumption–wealth ratio. Differentiating pC with respect to l and using
Eq. (3) we obtain
dpC 1
vÞ1g .
¼ ½1  ð1  db (4)
dl g
As shown in Eq. (4), terrorism increases the consumption–wealth ratio if go1 and
decreases the consumption–wealth ratio if g41. The reason is that terrorism reduces the
average return to investment and increases its variance. As a result, terrorism induces a
negative income effect and a positive substitution effect on consumption. The negative
income effect dominates when g41. However, the positive substitution effect dominates
for individuals with risk aversion smaller than that given by logarithmic utility.
(The substitution and income effects are derived in the Appendix.)
Finally, and most importantly for the purpose of this article, terrorism affects the
allocation of productive capital across countries. The international investment position of
the domestic economy is determined by the fraction of the world’s capital owned by
residents of the domestic economy, y, and the fraction of the world’s capital allocated to
production in the domestic economy, b v. In the notation of the model, the international
investment position of the domestic economy is equal to foreign holdings of domestic
vKðtÞ minus domestic holdings of foreign capital yð1  b
capital ð1  yÞb vÞKðtÞ. Therefore, the
international investment position of the domestic economy (normalized by the amount of
productive capital allocated to the domestic economy, b vKðtÞ) is equal to 1  y=b
v.
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 7

To investigate the effect of terrorism on net foreign investment, we differentiate Eq. (3)
with respect to l:
v
db vÞg
dð1  db
¼ 2 2
o0.
dl gðsW þ s2 vÞg1 þ l d2 gð1  d ð1  b
W Þ þ ld gð1  db vÞÞg1
Last equation shows that, in the model, terrorism has an unambiguously negative effect on
b
v. Notably, the magnitude of this effect is unbounded. In this simple two-country model,
the effect of terrorism on capital allocation across countries will be small if the direct
impact of terrorist attacks, represented by d, is small, as long as the degree of risk aversion
of international investors, g, is relatively large. However, if international investors are close
to risk neutrality (if g is close to zero), terrorist risk will have a large effect on the allocation
of capital across countries. The reason is that, in contrast to smooth risk, an increase in the
intensity of catastrophic terrorism not only increases the variance of the return to
investment, it also decreases its average. Investors with low levels of risk aversion have no
reason to diversify country risk, and react abruptly to relative changes in the intensity of
terrorism.
This may be an important consideration in practice. If international investors are
sufficiently diversified, they will have no reason to invest in countries with relatively high
levels of terrorist risk (if it is difficult to diversify terrorist risk locally). To illustrate this
point, suppose that the world economy consists of N economies (the domestic economy
plus N  1 foreign countries). To simplify the exposition, assume that only the domestic
economy is exposed to terrorism and that in the absence of a terrorist shock production in
country i (i ¼ 1; . . . ; N) is given by the stochastic process
avi ðtÞKðtÞ þ svi ðtÞKðtÞ dW i ðtÞ,
where W 1 ; . . . ; W N are independent Wiener processes. As before, terrorism in the domestic
economy is described by a Poisson process with coefficients ðl; dÞ. In this scenario, the
fraction of world’s capital invested in the domestic economy, b v, is given by
 
Nb v1
gs2 vÞg ¼ 0.
 ldð1  db
N 1
If there is no terrorism (l ¼ 0), then the domestic economy receives a fraction 1=N of
world’s capital. If there is terrorism in the domestic economy but not in the rest of the
world (l40), then b v will be smaller than 1=N. Moreover, notice that b v=ð1=NÞp
maxf1  ðld=gs2 ÞðN  1Þ; 0g, so for any given value of g, the ratio b v=ð1=NÞ will be small
when the number of countries, N, is large. The reason is that when investment can be
placed in many countries international investors are able to diversify risk without
allocating capital to countries with a higher relative risk of terrorism and therefore with a
lower expected return.
Fig. 1 shows how diversification opportunities accentuate the impact of terrorism on net
foreign investments. The left-hand side panel of Fig. 1 shows the value of b v=ð1=NÞ as a
function of the number of countries N, for g ¼ 1. The three series on the graph represent
three different values (100, 200, and 400) for s=ðl1=2 dÞ, which is the ratio of the standard
deviation of non-terrorist risk over the standard deviation of terrorist risk. The values of l
and d are set to 0.10 and 0.0005, respectively. The right-hand side panel of Fig. 1 shows the
same graph for g ¼ 10, a substantially higher degree of risk aversion. The value of b v=ð1=NÞ
decreases rapidly with N in all cases. In the case of g ¼ 1 and s=ðl1=2 dÞ ¼ 100, there is no
 ARTICLE IN PRESS
8 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

γ=1 γ = 10

Fraction of world′s capital in domestic economy/(1/N)

Fraction of world′s capital in domestic economy/(1/N)

1 1

0.8 0.8

0.6 0.6

0.4 0.4

0.2 0.2

0 0

0 20 40 60 80 0 20 40 60 80
Number of countries (N) Number of countries (N)

σ/(λ1/2δ) = 100

σ/(λ1/2δ) = 200

σ/(λ1/2δ) = 400

Fig. 1. Diversification and the effect of terrorism on investment.

investment in the domestic economy if the world consists of more than a few countries.
Even in the case of g ¼ 10 and s=ðl1=2 dÞ ¼ 400, b v=ð1=NÞ is about 10% lower than one
(the value that it would take in the absence of terrorism) with 80 countries. Fig. 1 shows
that, even if terrorist risk is only a small fraction of total economic risk, it may still have a
large economic impact in an open economy.
Rather than trying to capture all the channels through which terrorism may affect
economic outcomes, the simple model presented in this section emphasizes that the
diversification opportunities that arise in an integrated world economy can greatly amplify
the economic impact of terrorism (and, more generally, the impact of any one-sided risk).
This result is important because it suggests that, in an increasingly globalized world
economy, the ‘‘open-economy channel’’ may explain much of the difference between the
direct and the equilibrium impact of terrorism.

2.3. Model interpretation and empirical design

The equilibrium portrayed by our model can be interpreted as a description of the long-
run relationship between terrorism risk and international investment positions at the
country level. Because the model lacks transitional dynamics, it cannot effectively describe
short-run capital movements in response to variations in the intensity of terrorism.
Moreover, because diminishing returns to capital and adjustment costs are expected to be
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 9

substantial in the short run, a study of the effects of changes in the intensity of terrorism on
short-run capital movements will underestimate the long-run impact of terrorism.
Therefore, we choose to investigate the long-run effect of terrorism on international
investment positions rather than on capital flows.
We restrict the empirical analysis to direct investment, and do not consider debt and
portfolio investment. The reason is that debt and portfolio investment are heavily affected
by financial crises, the degree of development of the countries’ financial institutions, and
complicated capital control schemes.
The empirical evidence is based on a cross-section of countries. While measures of
terrorism risk in the cross-section of countries exist, we currently lack adequate
longitudinal risk data to conduct a direct study of capital flows. In principle, short-term
variations in terrorist risk levels could be used in a fixed-effects model to estimate the effect
of terrorism on FDI while controlling for unobserved determinants of FDI that are fixed in
the short run. In practice, however, the use of longitudinal data on terrorist risk would be
extremely problematic. On the one hand terrorist risk does not seem to display much short-
term variation.5 On the other hand, it may be difficult to establish the exact timing of the
impact of a change in terrorist risk on FDI. International investors may perceive short-
term changes in terrorist risk before or after those changes are reflected in measures of
terrorist risk, and the response of FDI to short-term changes in the level of terrorist risk
may be attenuated in the short run by the irreversible nature of direct investment.
Finally, notice that short-term longitudinal data on terrorist events or casualties may not
be adequate to estimate the impact of terrorism risk in a fixed effects model. The reason is
that, in most countries, terrorist incidents are low-probability events. Therefore, short-
term variation in the level of terrorist activity may be large even if the underlying level of
terrorist risk experiences little variation. In other words, terrorist events or casualties are
noisy measures of the latent level of terrorist risk. As a result, the regression coefficients on
variables measuring terrorist events or casualties will be biased towards zero. Time
differencing the data in a fixed effects model is known to exacerbate errors-in-variables
biases (Griliches and Hausman, 1986). In addition, the use of direct indicators of terrorist
events or casualties present other problems, which are described in Section 3.
Of course, the main disadvantage of using cross-sectional data is the potential for
omitted variable bias. To address this problem we perform a careful sensitivity analysis
and demonstrate that the results of the empirical section are robust to many different
specifications.

3. The data

The model we developed in the previous section predicts that, even if terrorist attacks
destroy only a small fraction of the productive capital of a country, increases in terrorist
risk may greatly affect the allocation of international investment across countries. In the
next section, we use a unique data set on terrorist and other country risks to quantify the
impact of terrorism on net foreign direct investment (net FDI) positions.
The evidence presented below uses data on the net stock of FDI from a cross-section of
countries. Data on FDI stocks come from the United Nations Conference on Trade And
5
For example, the terrorist risk index produced every year since 1995 by the Economist Intelligence Unit for a
set of 60 countries did not report any change in the level of terrorist threat for any country until the year 2001.
 ARTICLE IN PRESS
10 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

Development (UNCTAD, 2004). The UNCTAD database provides information on FDI

stocks in the year 2003 for 196 countries and territories. Net FDI positions are measured as
direct investment liabilities minus direct investment assets normalized by GDP.6 We will
sometimes refer to the net FDI position as a fraction of GDP simply as net FDI positions.
Our measure of country terrorism risk is the World Markets Research Centre’s Global
Terrorism Index (GTI). The GTI seems to be the first comprehensive attempt to measure
globally the risk from terrorist attacks at the country level. The GTI covers 186 countries
and territories for the period 2003/2004. The GTI combines expert ratings of the
motivation, presence, scale, efficacy, and prevention of terrorism at the country level.7
Previous empirical studies on terrorism have used two types of indicators of terrorist
activity: the number of terrorist incidents and the number of casualties. These measures
may suffer from several drawbacks. As noted by Frey et al. (2007), the number of attacks
and casualties collected from official statistics or from the media may underestimate the
actual number of casualties caused by terrorism. In addition, data on terrorist incidents
group together attacks of different magnitude and significance. Perhaps more importantly,
in the risk-based analysis of this paper, terrorism is not simply a history of violence, but a
latent variable representing the uncertainty created by the possibility of future terrorist
attacks. Statistics on the number of terrorist events or casualties may not fully reflect the
level of terrorism perceived by international investors.
Terrorist risk ratings, like the GTI, do not suffer from these drawbacks and incorporate
information that relate to risk but are not reflected in casualties, like motivation of
terrorists or prevention by the authorities. Moreover, because risk ratings are commonly
used by international investors to evaluate specific country risks, they have the advantage
of directly reflecting one of the most important channels through which information about
terrorism risk is revealed to international investors. Consequently, terrorism risk ratings
are particularly relevant as a measure of terrorism risk to evaluate the effect of terrorism
on international investment. A disadvantage of terrorism risk ratings is that they provide
only a summary measure of an intrinsically complex phenomenon.
Like any other risk, terrorism should depress the stock of international investment in
any particular country. Therefore, empirical evidence on the ‘‘open-economy channel’’ of
terrorism can be obtained using a cross-section of countries. In our regressions, we
measure the amount of variance of net FDI stock (normalized by the country GDP)
explained by terrorist risk. Because international investors take into account other types of
risk factors such as the overall political, legal, and security environments of the target
country, this exercise must acknowledge sources of risk other than terrorism. As we will
show below, these other risk factors correlate with terrorist risk. Therefore, it is important
to control in our regressions for other risk factors which may also affect foreign
investment. Risk rating data allow us to do exactly this. We include in our regressions a
Country Risk Index (CRI) also produced by the World Markets Research Centre
(WMRC). The WMRC CRI combines six risk factors (political, economic, legal, tax,
operational, and security) into an overall CRI.

6
FDI assets (liabilities) are defined as the value of capital and reserves, including retained profits, invested in a
foreign (the domestic) economy attributable to the parent enterprise, plus the net indebtedness of affiliates to the
parent enterprise resident in the domestic (a foreign) economy. Data on FDI stocks are at book value or historical
cost, reflecting prices at the time when the investment was made.
7
See World Markets Research Center (2003) for further information on the GTI.
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 11

As explained in the previous section, the net foreign investment position of a country is
determined in part by the fraction of world’s productive capital owned by the residents of
that country. Because rich countries have higher per capita levels of capital, we include in
our regressions the countries’ levels of GDP per capita as an explanatory variable
(these data come from the World Bank, 2004).
Of course, the stock of foreign investment of a country depends crucially on the degree
of the country’s openness to capital mobility. To measure the degree of openness of a
country to foreign capital we use the index of capital flows and foreign investment
restrictions published as a factor of the Index of Economic Freedom by The Heritage
Foundation and The Wall Street Journal (Miles et al., 2004).8
Finally, regional factors not accounted for by the country risk ratings or the other
explanatory variables may also influence the investment position of a country. To control
for these effects, we include regional dummies in many of the regressions reported below.

4. Empirical analysis

4.1. Main results

This section assesses the effect of terrorism on FDI positions in an open world economy.
Some of the 186 countries included in the WMRC GTI impose severe restrictions on FDI,
and hence they do not comply with the open market environment considered in this article.
For this reason we use the classification published by the International Monetary Fund
(IMF, 2001b) to restrict our sample to 110 countries with no restrictions on FDI
repatriations, and available data on FDI assets and liabilities as well as GDP.
Additionally, when a measure of openness to FDI is included in the regression the sample
is further restricted to 98 countries, because of data availability. Table B1 in Appendix B
lists the countries in the full 186-country sample and subsamples.
Table 1 describes the terrorist and country risk data for the 186-country full sample and
the 110-country regression sample. Among country risk factors, security exhibits the
highest correlation with terrorist risk (GTI). Part of the correlation between the terrorism
index and the country risk security factor is created mechanically, because the security
factor of the CRI incorporates a measure of terrorism in addition to other measures of civil
unrest, crime, and external security threats. We construct an alternative CRI, termed
corrected CRI, based on the same factors as the original CRI but excluding security risk
(see Appendix B). The cross-country regressions reported below use both the original and
the corrected CRI. Regressions with the original CRI should underestimate the negative
impact of terrorism on FDI, because terrorism risk is one of the components of that index.
Regressions with the corrected CRI should overestimate the negative impact of terrorism
on FDI, because the regression does not take into account the effect of other security risks
which are likely to have a positive correlation with terrorist risk. Using both the original
and the corrected CRI we obtain coefficients that can be interpreted, respectively, as a
lower bound and an upper bound to the negative effect of terrorism on net FDI positions.
As we show below, the regression results depend little on which CRI, corrected or
uncorrected, is used as a regressor.
8
Available at: http://www.heritage.org/research/features/index/.
 ARTICLE IN PRESS
12 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

Table 1
Country risk ratings

Mean SD Range Correlation matrix

GTI CRI Political Economic Legal Tax Operational

A. 186-country full sample

Terrorist risk
GTI 40.13 19.82 [10, 94]
Overall country risk
CRI 2.80 0.86 [1.15, 4.72] 0.2968
Country risk factors
Political 2.80 0.90 [1, 4.5] 0.2571 0.9564
Economic 2.91 0.88 [1, 4.75] 0.2289 0.9553 0.8950
Legal 2.66 0.98 [1, 5] 0.2154 0.9428 0.8656 0.8740
Tax 2.56 0.95 [1, 5] 0.2107 0.9020 0.8194 0.8139 0.8798
Operational 2.89 0.92 [1, 5] 0.2941 0.9461 0.8718 0.8872 0.9014 0.8626
Security 2.71 1.01 [1, 4.75] 0.5633 0.8523 0.8095 0.7731 0.7548 0.6935 0.7966
Country risk net of security risk
Corrected CRI 2.80 0.86 [1.16, 4.74] 0.2527

B. 110-country regression sample

Terrorist risk
GTI 40.23 18.77 [12.5, 84]
Overall country risk
CRI 2.67 0.83 [1.26, 4.32] 0.1868
Country risk factors
Political 2.69 0.90 [1, 4.5] 0.1212 0.9594
Economic 2.79 0.87 [1.25, 4.5] 0.1126 0.9581 0.8953
Legal 2.51 0.96 [1, 4.5] 0.1431 0.9514 0.8772 0.9011
Tax 2.43 0.82 [1, 4] 0.1220 0.9031 0.8435 0.8211 0.8825
Operational 2.73 0.88 [1, 4.25] 0.1920 0.9488 0.8841 0.8991 0.9041 0.8473
Security 2.63 1.01 [1, 4.5] 0.4946 0.8487 0.7934 0.7595 0.7635 0.6992 0.8135
Country risk net of security risk
Corrected CRI 2.67 0.84 [1.28, 4.32] 0.1364

Table 2 reports descriptive statistics for the economic variables. Net FDI positions may
represent a large quantity relative to GDP and exhibit high disparities among countries; in
our regression sample, net FDI positions over GDP average 28 percentage points with a
standard deviation of 39 percentage points. Net FDI positions and national income are
both negatively correlated with our measure of terrorist risk, whereas the level of FDI
restrictions has a positive correlation with terrorism. Table 2 also reports descriptive
statistics for demographic factors, governance measures and other explanatory variables
used in the regressions reported later in the article.
Table 3 reports cross-country regressions for the 110-country sample. Column (1)
reports the result of regressing net FDI positions (over GDP) on a constant term and the
GTI. The coefficient on the terrorism risk is negative and statistically different from zero at
conventional test levels. The R-squared coefficient indicates that terrorism explains about
2% of the cross-country variance in net FDI positions.
The coefficient on the GTI in column (2) remains negative and significant after including
log GDP per capita in the regression specification. As expected, log GDP per capita
exhibits also a negative regression coefficient. The results remain similar when we include
regional dummies in the regression specification in column (3).
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 13

Table 2
Descriptive statistics for economic variables

Number of Mean SD Range Correlation

Countries with GTI

Net FDI position (over GDP) 110 0.28 0.39 [0.62, 2.97] 0.1360
Log per capita GDP 110 7.91 1.61 [4.88, 10.79] 0.0639
FDI restrictions 98 2.66 0.91 [1,5] 0.0714
Demographic factors
Population under 14 (percentage) 109 29.22 10.43 [14.08, 48.88] 0.1045
Population over 65 (percentage) 109 8.23 4.97 [2.21, 18.85] 0.0953
Net primary enrollment rate 84 89.92 12.26 [33.96, 104.54] 0.1285
Governance factors
Voice and accountability 110 0.16 0.94 [2.05, 1.72] 0.1885
Political stability 110 0.14 0.95 [2.28, 1.63] 0.4750
Government effectiveness 110 0.13 1.03 [1.56, 2.26] 0.1051
Regulatory quality 110 0.21 0.95 [1.95, 1.93] 0.1259
Rule of law 110 0.13 1.01 [1.76, 2.03] 0.1808
Control of corruption 110 0.11 1.06 [1.70, 2.39] 0.1480
Macroeconomic factors
Credit (over GDP) 73 0.58 0.45 [0.04, 1.61] 0.1902
Real effective X-rate range 58 26.60 15.78 [6.97, 58.22] 0.0029
Gov. consumption (over GDP) 108 16.46 6.59 [5.40, 42.43] 0.2367
Std. Dev. growth rate 104 0.04 0.04 [0.00, 0.38] 0.0254
Other factors
Earthquake risk index 109 0.28 1.76 [0, 17.74] 0.0550

Table 3
Baseline specifications

Dependent variable Net FDI positions over GDP

(1) (2) (3) (4) (5)

Constant 0.3944** 1.0465** 0.6996 1.5791** 1.5354**

(0.0660) (0.2686) (0.4124) (0.4115) (0.4298)
Terrorist risk 0.0028** 0.0033** 0.0028* 0.0021* 0.0025**
(0.0014) (0.0013) (0.0015) (0.0012) (0.0012)
Log GDP per capita 0.0802** 0.0549 0.1125** 0.1088**
(0.0321) (0.0395) (0.0315) (0.0332)
FDI restrictions 0.0611* 0.0622*
(0.0326) (0.0339)
Country risk 0.1183**
(0.0571)
Corrected country risk 0.1047*
(0.0599)
Regional dummies No No Yes Yes Yes
Number of countries 110 110 110 98 98
R-squared 0.02 0.02 0.16 0.31 0.30

Robust standard errors in parentheses.

Regional dummies: middle East and North Africa, Latin America and the Caribbean, Eastern Europe and Central
Asia, Rest of Asia, Sub-Saharan Africa and Western Europe.
**
Indicates statistical significance at the 5% level.
*
Indicates statistical significance at the 10% level.
 ARTICLE IN PRESS
14 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

In column (4) we include other potential determinants of foreign investment: a measure

of the severity of the barriers to foreign capital and the WMRC CRI. As expected, the new
variables exhibit negative coefficients. The inclusion of the overall CRI and the other new
variables leaves the coefficient on the GTI with a value of 0:21% of GDP and significant
at the 10% level.
As explained above, because one component of the security factor of the CRI is a
measure of terrorism, the coefficient on the GTI in column (4) incorporates a positive bias
relative to a hypothetical regression using a measure of country risks which does not
account for terrorism. Therefore, the coefficient on the GTI in column (4) can be
interpreted as an upper bound on the coefficient of interest (that is, it underestimates the
negative impact of terrorism on net FDI). To obtain a lower bound, in column (5) we
estimate the same specification as in column (4) but using a corrected version of the CRI
that does not include the security risk factor. Because security threats other than terrorism,
which are unaccounted for in column (5), are likely to be positively correlated with
terrorism and affect FDI negatively, the coefficient on the GTI incorporates a negative bias
(it overestimates the negative impact of terrorism on net FDI). The value of the coefficient
on terrorism risk in column (5) is 0:25% of GDP, slightly lower than the value of the
same coefficient in column (4).
To give an idea of the magnitude of the estimated effect of terrorist risk, let us consider
the effect of a one standard deviation change in terrorist risk on the net FDI position.
A standard deviation change in terrorist risk is a change of 19.82 (18.77 in the 110-country
regression sample) in the GTI. This change is about the difference in terrorist risk between
Italy and the United States (the latter being more risky). Using the parameter estimated of
Table 3 column (5), this increase in terrorist risk induces a fall in the net FDI position of
about 5% of GDP. In monetary terms this amounts to 16 billion 2003 US dollars for the
average country (14.2 billion in the 110-country regression sample). This figure is large
suggesting that the open-economy effect of terrorism may be substantial.

4.2. Robustness analysis

In this section we perform several robustness checks on the specification of the

regression equation. We extend the regression equation to include demographic factors,
human capital, governance indicators, macroeconomic risk, financial factors, and other
variables that could potentially affect the net FDI positions.

4.2.1. Demographic factors

As a first robustness check we include in our regression demographic factors. The
population age-structure may affect FDI. Life-cycle theory suggests that countries with
young populations should be net debtors to the rest of the world. In addition,
multinational corporations engaging in FDI often target countries where the population
is young, so labor is an abundant production factor. Furthermore, the empirical evidence
in Lane and Milesi-Ferretti (2002) indicates that demographic factors exert a significant
effect on the net foreign assets position of countries, especially for the developing world.
In Table 4 we expand the set of explanatory variables by including demographic factors:
the fraction of population under age 14, and the fraction of the population over age 65.
The results of columns (1)–(4) indicate that our prior results are robust against the
inclusion of measures of the percentages of the population under age 14 and over age 65.
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 15

Table 4
Demographic factors

Dependent variable Net FDI positions over GDP

(1) (2) (3) (4) (5)

Constant 1.5343** 1.6677** 2.3874** 2.4245** 1.7870**

(0.4312) (0.6446) (0.7155) (0.6971) (0.7360)
Terrorist risk 0.0027** 0.0024** 0.0024** 0.0021* 0.0019
(0.0012) (0.0012) (0.0012) (0.0012) (0.0013)
Log GDP per capita 0.0910** 0.1177** 0.1192** 0.1226** 0.0819
(0.0415) (0.0476) (0.0459) (0.0440) (0.0514)
FDI restrictions 0.0637* 0.0620* 0.0649* 0.0639* 0.0551
(0.0342) (0.0345) (0.0351) (0.0340) (0.0404)
Country risk 0.1129*
(0.0575)
Corrected country risk 0.1066* 0.1032* 0.1000* 0.0222
(0.0606) (0.0585) (0.0595) (0.0768)
Population under 14 0.0025 0.0159* 0.0158* 0.0100
(0.0066) (0.0090) (0.0090) (0.0095)
Population over 65 0.0128 0.0336* 0.0335* 0.0263
(0.0134) (0.0187) (0.0187) (0.0214)
Net primary enrollment 0.0014
(0.0034)
Number of countries 98 98 98 98 80
R-squared 0.31 0.31 0.34 0.34 0.32

Robust standard errors in parentheses.

All regressions include regional dummies.
**
Indicates statistical significance at the 5% level.
*
Indicates statistical significance at the 10% level.

These age-range population variables exhibit statistically significant coefficients only when
included together, as in columns (3) and (4).9
Countries’ levels of human capital plausibly influence international investment
positions. In order to determine the importance of human capital in the determination
of net FDI positions we have included in column (5) of Table 4 a proxy for human capital,
the net enrollment rate in primary school. These data, from United Nations Educational
Scientific and Cultural Organization (UNESCO), were available for 80 countries of the
regression sample. We find that the coefficient on the enrollment rate is not statistically
significant. The inclusion of this variable in the regression increases the standard errors of
the other estimated coefficients and renders them insignificant (probably because of a
combination of the smaller sample size on the regression in column (5) and the high
correlation between enrollment rate and other included variables).10

9
The fractions of population under 14 and over 65 are negatively correlated in our sample, with correlation
coefficient equal to 0:9188. This strong negative correlation along with the negative regression coefficients on the
two variables implies that when one of them is omitted from the regression the coefficient on the included variable
will include a positive bias.
10
Using the average number of years of schooling of adults and the net enrollment rate in secondary school as
proxies for the level of human capital yields very similar results, but the sample is restricted to a smaller number of
countries.
 ARTICLE IN PRESS
16 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

4.2.2. Other types of country risks

While this paper focuses on terrorist risk, other types of country risk may be important
in international investment decisions. Here we test the robustness of our findings on
terrorist risk by including in our specification the factors that are used to calculate the
WMRC CRI.
The WMRC CRI used in our regressions accounts for political, economic, legal, tax,
operational, and security factors by aggregating them into a single measure of country risk.
The political factor measures political instability and indicates how mature and well-
established the political system is. The economic factor assesses the economic stance of the
country in terms of the freedom of market forces and the macroeconomic fundamentals of
the economy. The legal factor indicates how well-established the legal system is and
whether necessary business laws are in place. The tax factor measures the coherence and
fairness of the tax system as well as the overall level of taxes. The operational factor is an
assessment of the degree of openness to FDI, the quality of infrastructure, how well the
labor market functions, and bureaucracy and corruption. Finally, the security factor
measures threats originating from civil unrest, crime, terrorism, as well as external threats.
Table 5 reports regressions where we have included political, economic, legal, tax,
operational, and security factors. None of the first five factors significantly influence FDI
positions either taken together (column (1)) or one at a time (columns (2)–(6)). The
previous results for terrorist risk and the other explanatory variables remain robust across
these regressions. In column (7), we introduce in the regression the security factor of the
WMRC CRI (which includes terrorist risk) in lieu of the WMRC GTI. The coefficient on
the security factor is negative and statistically significant. The results in Table 5 suggest
that terrorism and, more generally, security are among the most apparent country-specific
determinants of the stock of net FDI.

4.2.3. Governance indicators

Here we test the robustness of our findings against potential omitted governance
variables that may be important determinants of net FDI positions. To control for
corruption and other forms of mis-governance, we employ a data set compiled by
Kaufmann et al. (2004) that provides indicators on the efficacy of six dimensions of
governance. These dimensions include:

1. Voice and accountability: measures an aggregate of civil liberties, political rights, and
other aspects of the political system representing the accountability of the government.
2. Political stability and absence of violence: measures the likelihood that the government
will not be destabilized or overthrown by violent means, including terrorism.
3. Government effectiveness: measures the ability of governments to implement and
enforce their policies.
4. Regulatory quality: measures the absence of market-unfriendly policies.
5. Rule of law: measures the protection of property rights and the extent to which residents
have confidence in the rules of society.
6. Control of corruption: measures the degree to which public power is not perceived by
the population as used for private gain.

The political stability indicator, which incorporates (lack of) terrorism, exhibits a high
correlation with the WMRC GTI in Table 2. We report the regression results including
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 17

Table 5
Country specific risks

Dependent variable Net FDI positions over GDP

(1) (2) (3) (4) (5) (6) (7)

Constant 2.4596** 2.1268** 2.3859** 2.1799** 2.0688** 2.3450** 2.1225**

(0.8291) (0.6562) (0.7351) (0.6752) (0.63701) (0.7528) (0.6094)
Terrorist risk 0.0024* 0.0025** 0.0025** 0.0026** 0.0027** 0.0023*
(0.0014) (0.0012) (0.0012) (0.0012) (0.0012) (0.0012)
Log GDP per capita 0.1252** 0.0960** 0.1192** 0.1022** 0.0976** 0.1124** 0.1035**
(0.0529) (0.0404) (0.0464) (0.0407) (0.0376) (0.0519) (0.0340)
FDI restrictions 0.0668* 0.0750** 0.0676* 0.0734** 0.0728** 0.0744** 0.0768**
(0.0356) (0.0356) (0.0351) (0.0339) (0.0343) (0.0350) (0.0316)
Population under 14 0.0160 0.0164* 0.0161* 0.0167* 0.0149 0.0168* 0.0157*
(0.0102) (0.0091) (0.0089) (0.0092) (0.0096) (0.0091) (0.0096)
Population over 65 0.0325* 0.3459* 0.0321* 0.0345* 0.0319* 0.0358* 0.0328*
(0.0181) (0.0186) (0.0187) (0.0188) (0.0185) (0.0183) (0.0184)
Political risk 0.0103 0.0413
(0.0624) (0.0461)
Economic risk 0.0802 0.0910
(0.0767) (0.0554)
Legal risk 0.0074 0.0475
(0.0677) (0.0431)
Tax risk 0.0105 0.0481
(0.1077) (0.0588)
Operational risk 0.0311 0.0684
(0.0993) (0.0635)
Security risk 0.0846**
(0.0301)
Number of countries 98 98 98 98 98 98 98
R-squared 0.34 0.33 0.34 0.33 0.33 0.32 0.34

Robust standard errors in parentheses.

All regressions include regional dummies.
**
Indicates statistical significance at the 5% level.
*
Indicates statistical significance at the 10% level.

these six indicators of governance in columns (1)–(6) of Table 3. None of the coefficients of
these indicators are statistically significant. More importantly, with the exception of
column (2) the regression coefficient on the terrorism risk variable remains virtually
unchanged through these regressions. The relative change of the coefficient on terrorism
risk in column (2) reflects the mechanical relationship between the index of political
stability and terrorism. In addition, the inclusion of the political stability index in the
regression increases the standard error of the regression coefficient on the terrorism index
variable, due to the high correlation between the political stability index and the GTI. The
results in Table 6 indicate that, when we include the corrected CRI and the other
previously tested variables in our regression, additional indicators of quality of governance
do not have a significant effect on net FDI positions.

4.2.4. Financial factors, macroeconomic risk, and natural disasters

International capital flows have been found to depend on the degree of financial
development in a country (see, e.g., Albuquerque et al., 2005). In a general sense, stock
 ARTICLE IN PRESS
18 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

Table 6
Governance indicators

Dependent variable Net FDI positions over GDP

(1) (2) (3) (4) (5) (6)

Constant 2.5294** 2.3439** 2.4514** 2.4396** 2.4062** 2.3516**

(0.7165) (0.7665) (0.7178) (0.7750) (0.7184) (0.7318)
Terrorist risk 0.0027** 0.0019 0.0026** 0.0025** 0.0026* 0.0027*
(0.0011) (0.0022) (0.0013) (0.0012) (0.0015) (0.0014)
Log GDP per capita 0.1164** 0.1204** 0.1119* 0.1166** 0.1169** 0.1074*
(0.0470) (0.0460) (0.0566) (0.0453) (0.0518) (0.0572)
FDI restrictions 0.0740* 0.0635* 0.0695 0.0709* 0.0658* 0.0695*
(0.0357) (0.0350) (0.0420) (0.0411) (0.0374) (0.0390)
Corrected country risk 0.1501** 0.0862 0.1362 0.1237 0.1091 0.1183**
(0.0670) (0.0771) (0.0961) (0.0944) (0.0730) (0.0589)
Population under 14 0.0160* 0.0162* 0.0155 0.0155* 0.0159* 0.0149*
(0.0900) (0.0090) (0.0094) (0.0089) (0.0091) (0.0094)
Population over 65 0.0299 0.3453* 0.0331* 0.0328* 0.0335* 0.0319*
(0.0187) (0.0182) (0.0185) (0.0182) (0.0188) (0.0183)
Governance factors
Voice and accountability 0.0796
(0.0620)
Political stability 0.0248
(0.0709)
Government effectiveness 0.0467
(0.1268)
Regulatory quality 0.0307
(0.0811)
Rule of law 0.0135
(0.0839)
Control of corruption 0.0390
(0.0746)
Number of countries 98 98 98 98 98 98
R-squared 0.35 0.34 0.34 0.34 0.34 0.34

Robust standard errors in parentheses.

All regressions include regional dummies.
**
Indicates statistical significance at the 5% level.
*
Indicates statistical significance at the 10% level.

markets facilitate the flow of a significant share of FDI. Furthermore, newly established
foreign firms, both those acquired through the stock market and those established as FDIs,
require financial services. Thus, as they make investment decisions foreign investors should
account for the degree of development of a country’s financial sector, and particularly the
development of the banking sector, as both of these factors will affect the returns of their
investments. The first column of Table 7 reports a regression where we include a common
measure of financial development, the total volume of credit of the banking industry as a
share of the GDP, as an additional explanatory variable. This measure of financial
development does not exhibit a statistically significant coefficient and its inclusion renders
most of the other regressors insignificant; the coefficient on the terrorism risk variable
remains, however, marginally significant.
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 19

Table 7
Financial factors, macroeconomic risk, and natural disasters

Dependent variable Net FDI positions over GDP

(1) (2) (3) (4) (5)

Constant 1.7769** 2.5409* 2.3820** 2.3129** 2.3275**

(0.7285) (1.4539) (0.7017) (0.7270) (0.7148)
Terrorist risk 0.0023* 0.0021 0.0025** 0.0025** 0.0024**
(0.0014) (0.0019) (0.0012) (0.0012) (0.0012)
Log GDP per capita 0.0593 0.0658 0.1192** 0.1130** 0.1156**
(0.0682) (0.1024) (0.0484) (0.0484) (0.0459)
FDI restrictions 0.0806* 0.0965* 0.0570 0.0629* 0.0676*
(0.0440) (0.0583) (0.0361) (0.0355) (0.0356)
Corrected country risk 0.0989 0.0306 0.1131* 0.1045 0.0926
(0.1100) (0.1320) (0.0594) (0.0646) (0.0601)
Population under 14 0.0097 0.0336* 0.0162* 0.0157 0.0155*
0.0095 (0.0192) (0.0093) (0.0095) (0.0090)
Population over 65 0.0251 0.0666** 0.0356* 0.0333* 0.0329*
(0.0205) (0.0264) (0.0186) (0.0184) (0.0188)
Credit (over GDP) 0.1859
(0.1423)
Real eff. X-rate range 0.0028
(0.0031)
Gov. cons. (over GDP) 0.0021
(0.0063)
Std. dev. growth rate 0.3841
(1.1798)
Earthquake risk index 0.0092*
(0.0049)
Number of countries 71 55 97 95 98
R-squared 0.38 0.50 0.35 0.34 0.34

Robust standard errors in parentheses.

All regressions include regional dummies.
**
Indicates statistical significance at the 5% level.
*
Indicates statistical significance at the 10% level.

Exchange rate fluctuations affect international capital flows and stocks of foreign
investment. Risk averse firms may decide not to invest in a country if exchange rate
volatility is high (see, e.g., Campa, 1993). We have constructed a measure of foreign
exchange fluctuations as the range of variation of the real effective exchange rate over the
1994–2003 period. Data availability for the construction of this variable restricts the
sample to 55 countries only. The second column of Table 7 includes this measure of
exchange rate fluctuations as an additional regressor. The evidence for this subsample
indicates that this financial factor does not exhibit a significant coefficient and its inclusion
renders the coefficient on the other regressors insignificant, potentially an effect of the
small sample size.
It has been documented that a high degree of intervention of the government in the
economy may deter foreign investment (see, e.g., Albuquerque et al., 2005). Using the ratio
of government consumption over GDP as a measure of government intervention, we
obtain the results reported in column (3) of Table 7. The coefficient on the additional
 ARTICLE IN PRESS
20 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

regressor is not statistically significant. The inclusion of the measure of government

intervention does not affect greatly the coefficients on the terrorism risk index and other
variables, or their statistical significance.
Growth instability potentially affects FDI as a source of uncertainty (see, e.g.,
Albuquerque et al., 2005). While a fast-growing economy might attract foreign investment,
macroeconomic instability might deter foreign investment. To account for growth
variability we use the standard deviation of GDP growth over the decade 1994–2003.
Column (4) of Table 7 indicates that growth variability is not a significant determinant of
the net foreign investment position. However, the coefficients on terrorist risk and the
other explanatory variables remain significant.
The theoretical model developed in this article conceptualizes terrorism as one-sided
catastrophic risk. However, terrorism is not the only instance of one-sided catastrophic risk.
To further validate our empirical findings, we include in column (5) of Table 7 a measure of
earthquake risk (the number of deaths caused by earthquakes during the period 1994–2003
normalized by 100,000 population in 2003).11 If our empirical approach to investigate the
effects of catastrophic risk on investment is appropriate, this new variable should exhibit a
negative coefficient in our regression. Indeed, the coefficient on the earthquake risk variable
is negative and significant at the 10% level. The inclusion of this variable in the regression
leaves the coefficient of terrorist risk virtually unchanged.

4.3. Discussion

Our results demonstrate that terrorism risk has significant explanatory power on net FDI
investment positions. In principle, the impact of terrorism on FDI positions estimated in our
regressions would be exacerbated if omitted determinants of FDI are negatively correlated
with terrorism. We show, however, that our regression results are robust under alternative
specifications, which control for a variety of factors suspected to affect foreign investment.
On the contrary, there are a number of reasons to believe that our estimates may be, in
fact, conservative.
First, our estimates of the impact of terrorism on foreign investment will be biased by reverse
causation if the presence of foreign capital in a country induces a terrorist response. Notice how-
ever that, in this case, reverse causation would create a positive bias in the estimated coefficient
on terrorism risk. Because we estimate a negative coefficient on the terrorism risk variable, the
potential bias created by reverse causation would not change the qualitative conclusions of this
study, i.e., the true coefficient would also be negative but larger in absolute value.
Moreover, notice that most of the estimates in this section reflect the degree of statistical
association between terrorism and net FDI positions over GDP holding constant the level
of per-capita GDP of a country (and other explanatory variables). Therefore, our analysis
will underestimate the long-run effects of a terrorist shock on an economy if foreign
ownership of the capital of a country boosts the level of per-capita income of the country
(e.g., via transfers of technology), and if increases in per-capita income tend to reduce the
level of terrorism. Under these conditions the negative impact of a terrorism shock would
be exacerbated by feedback effects.12
11
Natural disaster data are from the OFDA/CRED International Disasters Data Base.
12
However, empirical studies have failed to find evidence of the existence of a causal effect of per-capita income
on terrorism. See Krueger and Laitin (2003) and Abadie (2006).
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 21

There are two additional reasons to believe that our estimates provide in fact a
conservative measure of the impact of terrorism on FDI. First, the country latent level of
terrorist risk cannot be directly observed and the GTI will measure it with error. If terrorism
measurement error is ‘‘classical’’ (additive and independent of the latent terrorist risk) our
estimates will be biased towards zero. Second, our estimates of the effects of terrorism on
capital flows are attenuated to the extend that they reflect only the impact on capital flows
across countries. If terrorist risk levels vary within countries, the effect of terrorism on the
regions that are most affected by it will be more severe than what our estimates suggest.

5. Conclusions

In this paper we have shown how terrorism influences the equilibrium decisions of
international investors in an integrated world economy. We have introduced terrorism as
catastrophic risk in a standard endogenous growth model and analyzed the effect of an
increase in terrorist risk on the net FDI position of countries. The model suggests that in
an integrated world economy, where international investors are able to diversify other
country risks, terrorism may induce large movements of capital across countries. The
empirical evidence, based on cross-country regressions, indicates that terrorist risk
depresses net foreign investment positions. This relationship is robust to the introduction
of demographic factors, country-specific risk indexes, governance indicators, and other
financial and macroeconomic factors such as per capita GDP and FDI restrictions which
might determine the country’s FDI position.
Our estimates suggest that a one standard deviation increase in the intensity of terrorism
produces a 5% fall in the net FDI position of the country (normalized by GDP). Both the
model and the empirical evidence suggest that the open-economy channel may be an
important avenue through which terrorism hurts the economy.

Acknowledgments

We thank Pol Antràs, Jeff Frankel, Dani Rodrik, Todd Sandler, Jaume Ventura, Andrés
Velasco, Richard Zeckhauser and seminar participants at the 2005 NBER Summer Institute
and 2006 ASSA Meetings for very useful comments. Erik Garrison provided expert research
assistance. Financial support for this research was generously provided through NSF Grant
SES-0350645 (Abadie), Spanish Ministry of Science and Technology Grant SEC2003-04826
and University of the Basque Country Grant 35.321-13511 (Gardeazabal).

Appendix A. Solution to the model

For the model in Section 2, the value function is

Z 1  
bs CðsÞ
1g
 1 
V ðk; tÞ ¼ max E e dsKðtÞ ¼ k ,
t 1g
where the maximum is taken over all feasible consumption plans. We impose the
transversality condition
lim E½V ðKðtÞ; tÞ ¼ 0.
t!1
 ARTICLE IN PRESS
22 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

Notice that V ðkÞ ¼ ebt V ðk; tÞ does not depend on t. Using this property of the value
function the Hamilton–Jacobi–Bellman (HJB) equation simplifies to

bV ðkÞ

c1g  1 qV ðkÞ 1 q2 V ðkÞ 2 2 2 2 2
¼ max þ ðavk þ a ð1  vÞk  cÞ þ ðsW v k þ s2
W ð1  vÞ k Þ
c;v 1g qk 2 qk2

þ lðV ðk  dvkÞ  V ðkÞÞ þ l ðV ðk  d ð1  vÞkÞ  VðkÞÞ .

Solving the maximization problem on the left-hand side of last equation, we obtain the
following first-order conditions (FOC):
qV ðkÞ
cg 
b ¼ 0,
qk

qV ðkÞ q2 V ðkÞ 2 2
ða  a Þk þ vk  s2
ðsW b W ð1  bvÞk2 Þ
qk qk2
qV ðk  dbvkÞ qV ðk  d ð1  b
vÞkÞ 
l dk þ l d k ¼ 0.
qk qk
The second-order conditions (SOC) are

cg1 o0,
gb

q2 V ðkÞ 2 2 2 2 q2 V ðkð1  db
vÞÞ 2 2 2 
 q V ðkð1  d ð1  b
vÞÞÞ 2 2
2
ðs W k þ s W k Þ þ l 2
d k þ l 2
d k o0.
qk qk qk

Try the solution

mk1g  1=b
V ðkÞ ¼ .
1g
For the FOC, we obtain

c ¼ m1=g k,
b

ða  a Þ  gðs2W b
v  s2
W ð1  b vÞg þ l d ð1  d ð1  b
vÞÞ  ldð1  db vÞÞg ¼ 0.

It is easy to see that the SOC hold if b vp1. In particular, the derivative of the
cX0 and 0pb
left-hand side of last equation with respect to b v is negative for 0pb vp1. Therefore,
evaluating the left-hand side of last equation at b v ¼ 0 and 1 implies the following
conditions for an interior solution:
   g
ða  a Þ þ gs2
W  ld þ l d ð1  d Þ 40,

ða  a Þ  gs2W  ldð1  dÞg þ l d o0.

 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 23

From the HJB equation, we obtain a value for m1=g which yields CðtÞ ¼ pC KðtÞ, where

1 1
pC ¼ v þ a ð1  b
b þ ðg  1Þðab v2 þ s2
vÞÞ  gðg  1Þðs2W b vÞ 2 Þ
W ð1  b
g 2

1g   1g
vÞ  1  l ½ð1  d ð1  b
 l½ð1  db vÞÞ  1 .

As shown in Section 2:

dpC 1 40 if go1;
vÞ1g 
¼ ½1  ð1  db
dl g p0 if gX1:

Terrorism exerts a substitution and an income effect on consumption. To obtain the

substitution effect, differentiate C with respect to l for a fixed level of lifetime utility
v0 (compensated by changes in Kð0Þ). We know that V ðKð0Þ; 0Þ ¼ V ðKð0ÞÞ ¼
ðmKð0Þ1g  1=bÞ=ð1  gÞ. Differentiating this expression for V ð0Þ ¼ v0 , we obtain
 
qKð0Þ 1 qm=ql g qpC =ql
¼ Kð0Þ ¼ Kð0ÞX0,
ql V ð0Þ¼v0 1g m 1  g pC

since m1=g ¼ pC . Evaluate the substitution effect at Kð0Þ,

   
dC dpC qKð0Þ 1 dpC
¼ Kð0Þ þ pC ¼ Kð0ÞX0.
dl V ð0Þ¼v0 dl ql V ð0Þ¼v0 1  g dl

The income effect is

 
dC dC g dpC
 ¼ Kð0Þp0.
dl dl V ð0Þ¼v0 1  g dl

Given the form of the value function, the transversality condition for this problem
becomes

lim E½ebt KðtÞ1g  ¼ 0.

t!1

To verify this condition we first apply Ito’s Lemma for jump-diffusions to obtain

v þ a ð1  b
d ln KðtÞ ¼ ðab v2 þ s2
vÞ  pC  12 ðs2W b vÞ2 ÞÞ dt
W ð1  b
 
v dW ðtÞ þ sW ð1  b
þ sW b vÞ dW ðtÞ
þ lnð1  dbvÞ dPðtÞ þ lnð1  d ð1  bvÞÞ dP ðtÞ.

Therefore, ln KðtÞ is also a jump-diffusion. Notice that

E½ebt KðtÞ1g  ¼ E½ebt eð1gÞ ln KðtÞ .

Because ln KðtÞ is also a jump-diffusion, we can apply the results in Duffie et al. (2000), to
obtain (under regularity conditions)

E½ebðtsÞ KðtÞ1g  ¼ eaðsÞþbðsÞ ln KðsÞ ,

 ARTICLE IN PRESS
24 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

with
qaðsÞ
v þ a ð1  b
¼ b þ ðg  1Þðab v2 þ s2
vÞ  pC Þ  12 gðg  1Þðs2W b vÞ 2 Þ
W ð1  b
qs
vÞ1g  1  l ½ð1  d ð1  b
 l½ð1  db vÞÞ1g  1
¼ gpC  ðg  1ÞpC ¼ pC ,

qbðsÞ
¼ 0,
qs

aðtÞ ¼ 0, and bðtÞ ¼ 1  g. As a result, making s ¼ 0, we obtain

E½ebt KðtÞ1g  ¼ K 1g

0 e
pC t
.

Therefore, the transversality condition holds if and only if pC 40.

Appendix B. Additional data information

 Foreign direct investment assets and liabilities, 196 countries and territories, 2003. The
United Nations Conference on Trade and Development.
 Global Terrorism Index, 186 countries and territories, 2003/2004. World Markets
Research Centre.
 Country Risk Index, 186 countries and territories, 2003. World Markets Research
Centre’s analysts rate the political, economic, legal, tax, operational, and security
environments of each country and combine those six factors into a single index given by
vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
#ffi
u"
u ð0:25  ðpolitical riskÞ2 Þ þ ð0:25  ðeconomic riskÞ2 Þ þ ð0:15  ðlegal riskÞ2 Þ
t .
ð0:15  ðtax riskÞ2 Þ þ ð0:10  ðoperational riskÞ2 Þ þ ð0:10  ðsecurity riskÞ2 Þ

 Corrected Country Risk Index (net of security risk):

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u " #
u 1 ð0:25  ðpolitical riskÞ2 Þ þ ð0:25  ðeconomic riskÞ2 Þ þ ð0:15  ðlegal riskÞ2 Þ
t .
0:9 ð0:15  ðtax riskÞ2 Þ þ ð0:10  ðoperational riskÞ2 Þ

 Gross Domestic Product in US dollars for 2003. The World Bank, World Development
Indicators.
 Population. The World Bank. World Development Indicators.
 Percentage of population aged 0–14 and 65 and over, 2002. The World Bank, World
Development Indicators.
 Index of capital flows and foreign investment restrictions, a factor of the Index of
Economic Freedom. The Heritage Foundation and The Wall Street Journal. Miles et al.
(2004), http://www.heritage.org/research/features/index/i.
 Restrictions on repatriations of FDI earnings (International Monetary Fund, 2001b).
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 25

 The net enrollment rate in primary school and the net enrollment rate in secondary
school. United Nations Educational Scientific and Cultural Organization (UNESCO).
 Average years of schooling of adults. The World Bank Education Statistics.
 Governance indicators (Kaufmann et al., 2004).
 Private credit by deposit money banks and other financial institutions to GDP.
Financial Structure and Economic Development Database, See Beck et al. (1999). This
database has been updated to 2001. http://www.worldbank.org/research/projects/
finstructure/database.htm.
 Real effective exchange rate. International Monetary Fund. International Financial
Statistics, series code (XXX..RECZF. . .). 1994–2003.
 General government final consumption expenditure (% of GDP). 1994–2003 average.
International Monetary Fund, International Financial Statistics.
 Standard deviation of growth rate. Computed as the standard deviation of the annual
growth rates of GDP (constant prices, local currency units) for the period 1994–2003.
International Monetary Fund. International Financial Statistics.
 Earthquake risk index. Computed as the number of deaths caused by earthquakes
during the years 1994–2003 (OFDA/CRED International Disasters Database, (http://
www.em-dat.net/) per 100,000 population in the year 2003.

Table B1
Country list

186-country sample
Afghanistan, Algeria, Andorra, Angola, Bahrain, Bangladesh, Barbados, Belgium, Belize, Benin, Bermuda,
Bhutan, Brazil, Brunei, Burkina Faso, Burundi, Cameroon, Cayman Islands, Central African Republic, Chad,
Chile, China, Colombia, Congo, Cote d’lvoire, Cuba, Cyprus, DRCongo, East Timor, Equatorial Guinea,
Ethiopia, Fiji, French Guiana, Gabon, Grenada, India, Iran, Iraq, Israel, Kazakhstan, Kuwait, Lao PDR, Libya,
Liechtenstein, Luxembourg, Macau, Madagascar, Mali, Malta, Martinique, Moldova, Mozambique, Myanmar
(Burma), Namibia, Nepal, Niger, Nigeria, North Korea, Oman, Puerto Rico, Qatar, Russia, Rwanda, Samoa,
Saudi Arabia, Senegal, Somalia, Sri Lanka, Suriname, Syria, Taiwan, Tanzania, Togo, United Arab Emirates,
Uzbekistan, Zimbabwe, plus those countries in the 110-country sample
110-country sample
Antigua and Barbuda, Comoros, Dominica, Eritrea, Liberia, Maldives, Palestinian Authority, Papua New
Guinea, Sao Tome, Serbia and Montenegro, Seychelles, Sudan, plus those countries in the 98-country sample
98-country sample
Albania, Argentina, Armenia, Australia, Austria, Azerbaijan, Bahamas, Belarus, Bolivia, Bosnia and
Herzegovina, Botswana, Bulgaria, Cambodia, Canada, Cape Verde, Costa Rica, Croatia, Czech Republic,
Denmark, Djibouti, Dominican Republic, Ecuador, Egypt, El Salvador, Estonia, Finland, France, Gambia,
Georgia, Germany, Ghana, Greece, Guatemala, Guinea, Guinea Bissau, Guyana, Haiti, Honduras, Hong Kong
SAR, Hungary, Iceland, Indonesia, Ireland, Italy, Jamaica, Japan, Jordan, Kenya, Kyrgyzstan, Latvia, Lebanon,
Lesotho, Lithuania, Macedonia, Malawi, Malaysia, Mauritania, Mauritius, Mexico, Mongolia, Morocco,
Netherlands, New Zealand, Nicaragua, Norway, Pakistan, Panama, Paraguay, Peru, Philippines, Poland,
Portugal, Romania, Sierra Leone, Singapore, Slovakia, Slovenia, South Africa, South Korea, Spain, Swaziland,
Sweden, Switzerland, Tajikistan, Thailand, Trinidad and Tobago, Tunisia, Turkey, Turkmenistan, Uganda,
Ukraine, United Kingdom, United States, Uruguay, Venezuela, Vietnam, Yemen and Zambia
 ARTICLE IN PRESS
26 A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27

References

Abadie, A., 2006. Poverty, political freedom, and the roots of terrorism. American Economic Review, Papers and
Proceedings 96 (2), 50–56.
Abadie, A., Gardeazabal, J., 2003. The economic cost of conflict: A case study of the Basque country. American
Economic Review 93, 113–132.
Albuquerque, R., Loayza, N., Servén, L., 2005. World market integration through the lens of foreign direct
investors. Journal of International Economics 66 (2), 267–295.
Beck, T., Demirguc-Kunt, A., Levine, R., 1999. A new database on financial development and structure.
The World Bank. Policy Research Working Paper No. WPS 2146.
Becker, G., Murphy, K., 2001. Prosperity will rise out of the ashes. Wall Street Journal October 29, 2001.
Becker, G., Rubinstein, Y., 2004. Fear and the response to terrorism: An economic analysis. Mimeo.
Blomberg, S.B., Mody, M., 2005. How severely does violence deter international investment? Department of
Economics, Claremont McKenna College, Claremont, CA.
Campa, J.M., 1993. Entry by foreign firms in the United States under exchange rate uncertainty. Review of
Economics and Statistics 75 (4), 614–622.
Chen, A.H., Siems, T.F., 2004. The effects of terrorism on global capital markets. European Journal of Political
Economy 20 (2), 349–366.
Duffie, D., Pan, J., Singleton, K., 2000. Transform analysis and asset pricing for affine jump-diffusions.
Econometrica 68 (6), 1343–1376.
Enders, W., Sandler, T., 1996. Terrorism and foreign direct investment in Spain and Greece. Kyklos 49 (3),
331–352.
Enders, W., Sandler, R.T., Parise, G.F., 1992. An econometric analysis of the impact of terrorism on tourism.
Kyklos 45, 531–554.
Enders, W., Sachsida, A., Sandler, T., 2006. The impact of transnational terrorism on US foreign direct
investment. Political Research Quarterly 59 (4), 517–531.
Frey, B.S., Luechinger, S., Stutzer, A., 2004. Valuing public goods: The life satisfaction approach. Working Paper
184. Institute for Empirical Research in Economics, University of Zurich.
Frey, B.S., Luechinger, S., Stutzer, A., 2007. Calculating tragedy: Assessing the costs of terrorism. Journal of
Economic Surveys 21 (1), 1–24.
Global Business Policy Council, 2004. FDI Confidence Index. A.T. Kearney, Alexandria, VA.
Griliches, Z., Hausman, J.A., 1986. Errors in variables in panel data. Journal of Econometrics 31, 93–118.
IMF, 2001a. How has September 11 influenced the global economy? World Economic Outlook.
IMF, 2001b. Annual report on exchange agreements and exchange restrictions. International Monetary
Fund.
Kaufmann, D., Kraay, A., Mastruzzi, M., 2004. Governance matters III: Governance indicators for 1996–2002.
World Bank Economic Review 18, 253–287.
Kraay, A., Ventura, J., 2000. Current accounts in debtor and creditor countries. Quarterly Journal of Economics
115 (4), 1137–1166.
Kraay, A., Ventura, J., 2002. Current accounts in the long and short run. NBER Macroeconomics Annual.
Krueger, A.B., Laitin, D.D., 2003. Kto Kogo?: A cross-country study of the origins and targets of terrorism.
Mimeo.
Lane, P.R., Milesi-Ferretti, G.M., 2002. Long-term capital movements. In: Bernanke, B.S., Rogoff, K. (Eds.),
NBER Macroeconomics Annual 2002, vol. 16. MIT Press, Cambridge, MA, London.
Li, D., 2002. Is the AK model still alive? The long-run relation between growth and investment re-examined.
Canadian Journal of Economics 35, 92–114.
McGrattan, E.R., 1998. A defense of AK models. Federal Reserve Bank of Minneapolis. Quarterly Review 22 (4),
13–27.
Miles, M.A., Feulner, E.J., Jr., O’Grady, M.A., Eiras, A.I., 2004. Index of economic freedom. The Heritage
Foundation/Wall Street Journal.
Obtsfeld, M., 1994. Risk taking, global diversification, and growth. American Economic Review 84 (5),
1310–1329.
OECD, 2001. Organisation for economic co-operation and development. Economic Outlook 71.
Pshisva, R., Suarez, G.A., 2006. Captive markets: The impact of kidnappings on corporate investment in
Colombia. Federal Reserve Research Paper No. 2006-18.
Turnovsky, S., 1997. International Macroeconomic Dynamics. MIT Press, Cambridge, MA.
 ARTICLE IN PRESS
A. Abadie, J. Gardeazabal / European Economic Review 52 (2008) 1–27 27

UNCTAD, 2002. United Nations Conference on Trade and Development. FDI downturn in 2001 touches almost
all regions press release TAD/INF/PR/36, 21 January 2002.
UNCTAD, 2004. United Nations Conference on Trade and Development. Foreign investment database hhttp://
www.unctad.orgi.
United States Congress, Joint Economic Committee, 2002. The Economic costs of terrorism.
World Market Research Center, 2003. Global Terrorism Index 2003/4.